Commit Graph

2467 Commits

Author SHA1 Message Date
Nick Marino
2d4e2b59ff Use a randomized key for the overload test
This helps ensure we get different preflists each time, and will
hopefully make the test more likely to uncover failures in the future.
2016-09-20 15:33:48 -04:00
Nick Marino
062115b2cc In overload test, filter specifically on get msgs 2016-09-20 14:43:32 -04:00
Nick Marino
74becb61a8 Remove unused run_queue_len function 2016-09-20 13:24:11 -04:00
Nick Marino
2445cad120 Simplify get_victim & fix intermittent w1c fails
There's no reason I can find to rule out any particular vnode when
choosing the victim. The way this function is used in this test, it
seems we're trying to make sure we pick a vnode on a different node from
the one we're connecting to, but I removed this check and forced it to
pick a vnode on the same node, and the test still passed fine.

This was also causing occasional issues with the w1c tests, because
there we use n=1; if we skip the first entry in the preflist because
it's on the excluded node, then we will pick a victim vnode that is not
actually being read from, and the test will fail.
2016-09-20 10:14:35 -04:00
Nick Marino
3d583db7d1 Correct spelling of "protection" in log message 2016-09-19 16:30:20 -04:00
Nick Marino
7b1ab3c961 Remove redundant wait_until in overload setup
This is not needed because we already have other checks in setup() that
will not pass until riak_kv is started up, and riak_kv will not start
until riak_core is done starting, and once riak_core is done starting we
can be guaranteed that get_chash_bin() will succeed. (When riak_core is
started, it calls riak_core_ring_manager:start_link from the sup, which
then calls init, which calls set_ring, which calls set_ring_global,
which inserts an entry into the ets table for chashbin.)
2016-09-19 13:22:01 -04:00
Nick Marino
8d0c9da04c Clean up overload test setup code slightly 2016-09-19 13:18:13 -04:00
Nick Marino
a62b13494c Merge pull request #1156 from basho/nem-revert-join-retry-changes
Revert changes to plan/commit retry logic
2016-09-19 11:34:59 -04:00
Nick Marino
f8003d6305 Revert "Always ensure nodes are ready before continuing"
This reverts commit 0d341f7210.
2016-09-19 10:47:31 -04:00
Nick Marino
97e080aceb Revert "Retry if we mysteriausly get nothing_planned"
This reverts commit 0e1dbc04bf.
2016-09-19 10:47:22 -04:00
Nick Marino
05fd551eab Revert "Remove redundant wait call"
This reverts commit 43a8978bab.
2016-09-19 10:47:10 -04:00
Brian Sparrow
e771fee2d8 Add support for AAE hashing version 0 (#1148)
* Add support for AAE hashing version 0

Additionally, change verify_aae to use 3 nodes for more complete
testing.

* Update repl_aae_fullsync.erl

* Update verify_aae.erl

* Do hashtree upgrade verification first

* Fix intercepts in aae_fullsync test

* Increase node count for more accurate test
2016-09-16 18:33:22 -04:00
Fred Dushin
41ed0c5409 Merge pull request #1155 from basho/fd-replication_object_reformat-fix
Used Nick's trick to use the previous riak.conf to start
2016-09-15 16:07:48 -04:00
Fred Dushin
e1e718fd65 Used Nick's trick to use the previous riak.conf to start (simulating a
downgrade from a previously upgraded riak.conf).
2016-09-15 14:13:21 -04:00
Fred Dushin
633b25efe1 Merge pull request #1154 from basho/fd-rtdev-upgrade-fix
Fix for replication_upgrade and replication2_upgrade tests
2016-09-15 11:29:14 -04:00
Nick Marino
cfcc3f7160 Merge pull request #1153 from basho/nem-improve-join_cluster-retry-logic
Improve join cluster retry logic
2016-09-15 11:21:37 -04:00
Fred Dushin
9ace9fff97 Changed call from rtdev:upgrade to rt:upgrade, which has the expected
signature
2016-09-15 10:28:00 -04:00
Nick Marino
43a8978bab Remove redundant wait call
This is no longer needed since we are guaranteed to call try_nodes_ready
at the end of every iteration of plan_and_commit.
2016-09-14 18:37:40 -04:00
Nick Marino
0e1dbc04bf Retry if we mysteriausly get nothing_planned
Not sure why we keep hitting this, but we shouldn't just assume the
plan was committed because testing has repeatedly shown otherwise.
2016-09-14 18:35:38 -04:00
Nick Marino
0d341f7210 Always ensure nodes are ready before continuing
The try_node_ready function would previously get called after the first
execution of plan_and_commit returned, and if it failed it would call
plan_and_commit again. However, once the second plan_and_commit call
finishes, try_node_ready would not get called again. This commit changes
this behavior so that try_node_ready is always called after
plan_and_commit succeeds.
2016-09-14 18:28:29 -04:00
Nick Marino
734f19666c Increase number of retries in try_nodes_ready
We seem to be timing out on this a lot, so waiting 1.5 seconds doesn't
seem like enough.
2016-09-14 16:00:03 -04:00
Nick Marino
70d98066c0 Add single-argument version of try_nodes_ready
This factors out the defaults timeouts into its own function, and will
simplify some other refactorings later on.
2016-09-14 15:58:25 -04:00
Fred Dushin
17932f3ef9 Merge pull request #1151 from basho/nem-fix-downgrade-via-config-copy
Add copy_conf function and fix verify_riak_object_reformat test
2016-09-14 15:14:31 -04:00
Nick Marino
8e5505d357 Throw an error if we try to use copy_conf in rtssh 2016-09-14 15:11:23 -04:00
Nick Marino
8f99422814 Call copy_conf via the rt module 2016-09-14 13:47:51 -04:00
Nick Marino
60eb47f71b Use copy_conf func to fix object reformat test 2016-09-14 11:15:26 -04:00
Nick Marino
a75a3811b8 Add rtdev:copy_conf function
This new utility function copies config from one devrel to another. This
is particularly useful when the first thing a test does is downgrade to
a previous release which doesn't support some new config option that's
been recently added. By copying the old config into the new devrel
before we start up the nodes, we can automatically avoid specifying any
new, incompatible config keys for that particular test.
2016-09-14 10:47:53 -04:00
Fred Dushin
a604a3186a Merge pull request #1147 from basho/update-yz-tests/zl/add-deletes-to-crdt-yz-2.2
Update yz tests/zl/add deletes to crdt yz 2.2
2016-09-13 17:01:51 -04:00
Doug Rohrer
a50f85e17b Merge pull request #1149 from basho/fd-debuglog-fix
Modified rtdev:get_node_debug_logs to return filename and file handle
2016-09-12 15:21:30 -04:00
Fred Dushin
396a6cf242 Modified rtdev:get_node_debug_logs to return filename and file handle
so that the debuglog archive can be uploaded.
2016-09-12 13:50:20 -04:00
Fred Dushin
d01e3b131b Merge pull request #1145 from basho/riak_search-cuttlefish
Changed expected default for riak_search to enabled
2016-09-12 13:18:18 -04:00
Zeeshan Lakhani
49344a0422 refactor test to include wait-for around validation due to yz changes to draining, settings 2016-09-09 14:03:05 -04:00
Fred Dushin
5b0582b772 Changed expected default for riak_search to enabled 2016-09-09 11:33:00 -04:00
Zeeshan Lakhani
9564ab5823 - add comments/logs around additional tests
- few formatting updates
2016-09-08 17:31:09 -04:00
Zeeshan Lakhani
6156b2a90c update intercepts for change to delete_operation/5 in yz branch->features+fixes/zl/handle-cleanup-allow_mult=true 2016-09-08 17:31:09 -04:00
Zeeshan Lakhani
b56d394b77 extend crdt test w/ partition/siblings/vtag test and make sure it behaves properly w/ new tombstone code in yz 2016-09-08 17:31:09 -04:00
Zeeshan Lakhani
6d69f28e5f handle deletes via aae cases 2016-09-08 17:31:09 -04:00
Zeeshan Lakhani
ecae32da10 test deletions w/ yz_crdt + tombstones + delete_mode -> keep 2016-09-08 17:31:09 -04:00
Nick Marino
f37f69007b Merge pull request #1006 from basho/newtest-zl-verify-dt-data-upgrade
add test to verify sets/maps data/meta/binary-info across mixed cluster/upgrade/downgrade a… [JIRA: RIAK-2778]
2016-09-08 17:29:49 -04:00
Zeeshan Lakhani
ab014b13f0 sure up connections after partition 2016-09-08 17:22:29 -04:00
Brett Hazen
42b12dd304 Merge pull request #1144 from basho/nem-fix-pb_security-disabled-search-failure
Disable use of job enable flags for this test
2016-09-08 14:04:53 -06:00
Zeeshan Lakhani
8fff405f2d since lts is 2.0.7, build 2.0.7 2016-09-08 14:26:19 -04:00
Zeeshan Lakhani
03045d2cb3 check caps and use lts 2016-09-08 14:26:01 -04:00
Nick Marino
beb20830ec Disable use of job enable flags for this test
If we leave it up to the Cuttlefish defaults, we end up with search 1.0
being disabled, which causes this test to fail.
2016-09-08 14:16:52 -04:00
Doug Rohrer
4d1b2f16ce Merge pull request #1139 from basho/fd-capabilities-cherrypick
Test capabilities being removed in a node downgrade scenario.
2016-09-08 13:05:55 -04:00
Fred Dushin
df149dad39 added pattern matching to actually make the test fail if riak_core is
not patched.
2016-09-08 12:03:33 -04:00
Nick Marino
76bf3c2ba3 Merge pull request #1143 from basho/nem-fix-overload-intercepts
Fix riak_kv_get_fsm intercept used by overload
2016-09-08 10:59:03 -04:00
Zeeshan Lakhani
5d081d9e4e add test to verify sets/maps across mixed cluster/upgrade/downgrade after sets->ordsets change 2016-09-08 10:39:48 -04:00
Nick Marino
f785ee78be Fix riak_kv_get_fsm intercept used by overload
The get FSM normally uses `start/4` instead of `start_link/4` now, which
caused the overload test to break. Fixed the intercept name and all is
working again now.
2016-09-07 16:57:24 -04:00
Jason Voegele
797db72779 Merge pull request #1142 from basho/ph-fix-turn_on_search_before_using
Fix "verify_search" by turning on search
2016-09-07 15:48:35 -04:00